Price Research Methods

Price research can be tricky. David Lyon, principal at Aurora Market Modeling did a great review of past and current price research methods in his tutorial “Research for Solid Price Decisions” at the 2010 Sawtooth Software Conference in Newport Beach, CA.

The recurring theme across his presentation was the need to take into account respondent psychology and infuse realism in how we ask pricing questions in order to get unbiased results. The price research methods reviewed include:

Direct Questioning

• Willingness-To-Pay Questions: These can get you a ballpark price for radically new products when the price range is unknown. Given their inherent bias, since these questions clearly focus on price and invite respondents to “lowball” and bargain, they should be used only for exploration and followed-up with other price research techniques.
• Buy-Response in Monadic Designs: This approach can work if done right, which means asking about purchase intent in the context of a product description without calling respondent’s attention to the price and thus avoiding the bias found in willingness-to-pay questions. However, in order to measure price sensitivity, we need at least two different monadic cells where different price points are asked. This is a simple and unbiased approach, but its precision depends on the sample size for each cell, which may be a problem for tight budgets since large samples are recommended and many cells/price points may need to be tested.
• Sequential Proto-Monadic: This approach which tests several price points in a sequential way within the same respondent doesn’t allow to disguise the focus on price and usually results in over-estimation of price sensitivity. David’s advice: Don’t bother using it.
• Van Westendorp Price Sensitivity Meter: With this approach, we look for “normal” and “acceptable” price ranges between price points considered too cheap or too expensive based on the idea that price acts as a proxy for quality. A big issue with PSM is that respondents tend to be inconsistent in how they answer the 4 questions included in the approach: At what price it is too cheap? At what price it is a good value? At what price it is getting expensive? At what price is too expensive? This method can be used for early exploration in small samples, but results should be taken with skepticism as we can expect “lowballing.”  The other problem with this approach is that it doesn’t measure purchase intent. However, even the improvement suggested by Newton-Miller-Smith in 1993 by adding purchase intent questions after the “good value” and the “getting expensive” questions, has the problem that we really don’t know the actual purchase probability unless we have market data to calibrate the results.

Trade-Off Methods

• Rating-Base Full Profile Conjoint: This method, where a full profile is presented with variations across all attributes, was widely used a few years back. The major problem with this approach is that it tends to understate price sensitivity.  A hypothesis as to why is that price is taken as an indicator of overall quality resulting in “reversed” price utilities (preference increases with price) or flattened utilities (not much change as price increase) for many respondents. David believes this is due to a lack of realism in the task, particularly when hypothetical products are presented, and respondents don’t have anything knowledge or experience to base their choice on, so they use price to evaluate the quality (if it is expensive it must be good).
• Discrete Choice Modeling or Choice-Based Conjoint (CBC):  This includes different types of designs:
  • Price-only Choice Models: In this method, respondents have to make a choice among the brand/price options shown, which a natural and easier task for respondents. Here, price is treated as a separate attribute for each brand.  It is a good general-purpose approach when concrete, realistic products are tested, although it doesn’t allow simulating the effects of adding or deleting products included in the original set (they are fixed and only prices vary).
  • Choices between hypothetical products: These are described by combinations of price and non-price attributes and levels. In order to be efficient, we need to focus on things respondents care about and model them accordingly. In order to get good results, we need to avoid impossible and illogical scenarios.

In the end, the key is to take into account respondent psychology in the price research methods you use. As David says “give them realistic, natural questions and you’ll get less biased answers.”